‘innocent’ compromised systems.43 The bot herder need not therefore be located geographically close to his C&Cs. Nonetheless, it is possible that local – particularly linguistic – links exist between perpetrators and some hosting providers, including so-called ‘bullet-proof’ providers.44 The section of this Chapter on ‘cybercrime perpetrators,’ for example, also notes the existence of perpetrator cybercrime hubs in Eastern Europe – corresponding with the pattern of high C&C rates in this subregion. Infected Figure 2.10: Botnet infections, by country (2010) computers (zombies) are the other half of the equation to C&Cs. Globally, at least seven million computer devices may be part of a Other botnet.45 estimates place the figure much higher.46 Figure Source: UNODC elaboration of Microsoft Security Intelligence Report. 2.10 shows approximate distribution of these infections, by country.47 per 100,000 MSRT executions <100 100-200 200-400 400-500 500-750 >750 Infection distribution shows a different pattern to that of C&Cs. Zombies cluster more heavily in Western Europe (as opposed to Eastern Europe for C&Cs), and show significant infection rates in North, Central and Southern America, as well as some countries in Eastern Asia. This distribution tends to represent countries with high numbers of active computer users. Estimates of total zombies and botnet size face significant limitations. Two important methodological distinctions affecting estimates include botnet ‘footprint’ versus ‘live population’,48 and measurement of zombie ‘IP addresses’ versus ‘unique devices.’49 In this respect, it should be noted that (due to methodological factors) the C&C estimate above concerns unique IP addresses, whereas the zombie estimate concerns computer devices. The two global figures are thus not easily comparable. 43 44 45 46 47 48 49 In addition, in more recent P2P botnets, any zombie computer can be a client or a server, precluding the need for a particular server for bots to download programs or receive instructions. With respect to hosting providers see, for example, HostExploit and Group IB, 2012. Top 50 Bad Hosts and Networks Report. UNODC calculations based on Microsoft, 2010. Microsoft Security Intelligence Report. Volume 9. Figure as of first half 2010. This estimate is of the same order of magnitude as that of Symantec, 2011. Internet Security Threat Report. 2011. Volume 17 (estimate of 4.5 million for 2010). See, for example, Acohido, B., 2010. Are there 6.8 million –or 24 million– botted PCs on the Internet? The Last Watchdog. Available at: http://lastwatchdog.com/6-8-million-24-million-botted-pcs-internet/ Zombies are plotted as bot infections identified per 100,000 runs of the Microsoft Malicious Software Removal Tool. Data from Microsoft, 2010. Microsoft Security Intelligence Report. Volume 9. The methodology covers only those machines running Windows update (approximately 600 million machines worldwide) and identifies only the most wide-spread bot infections. Nonetheless, independent methodologies find similar infection levels when calculated on an individual country basis. See, for example, van Eeten, M.J.G. et al., 2011. Internet Service Providers and Botnet Mitigation. A Fact-Finding Study on the Dutch Market. Faculty of Technology Police and Management, Delft University of Technology. Zombies join and leave botnets on a continuous basis as new machines are infected and existing zombies cleaned. In addition, infected machines may suffer from multiple infections or be temporarily migrated from one botnet to another (Abu Rajab, M., et al., 2007. My Botnet is Bigger than Yours (Maybe, Better than Yours): Why size estimates remain challenging). Proceedings of the first conference on first workshop on hot topics in understanding botnets. Berkeley, CA: Usenet Association. The botnet footprint refers to the aggregated total number of machines that have been compromised over time. The botnet live population denotes the number of compromised machines that are simultaneously connecting with a C&C server. A particular number of identified IP addresses does not usually correspond to the number of devices due to two network effects: (i) the short-term assignment of different IP addresses to the same device (DHCP ‘churn’), and (ii) the sharing of a single IP address by multiple devices (NAT). Depending upon the size of DHCP and NAT effects, the number of unique IP addresses may be greater or smaller than the corresponding number of actual devices. Due to high DHCP churn rates by commercial ISPs, the number of observed IP addresses is usually significantly greater than the number of devices. 34

Select target paragraph3

Connect to a paragraph
Connect to an entity
Disable highlights
Add to table of contents